The composite phenotype analysis identifies potential concerted responses of physiological systems to high altitude exposure

The composite phenotype analysis identifies potential concerted responses of physiological systems to high altitude exposure Yi Li 1,3,9,†, Meng Hao2,4,†, Zixin Hu2,10,†, Yanyun Ma2,3,†, Kun Wang2, Xiaoyu Liu1, Xianhong Yin1, Menghan Zhang 2, Yi Wang2, Meng Liang2, Yuan Guo2, Lei Bao2, Shixuan Zhang2, Shiguan Le1, Chenyuan Wu1, Dayan Sun1, Yang Wei1, Fei Wu2, Rui Zhang1,3, Lingxian Zhu2, Hui Zhang2, Shuai Jiang2,11, Xingdong Chen1,6, Xiaofeng Wang1, Yao Zhang7, Longli Kang7, Wenyuan Duan5, Bin Qiao5, Jiucun Wang1,3,8,∗ and Li Jin1,3,8,9,∗

Environmental stresses, such as temperature, disease and altitude could induce systematic changes of biological systems which manifests as concerted responses across multiple systems within a certain period of time [1]. High altitude acclimatization (HAA) refers to a series of adaptive physiological responses to hypoxic stress. During these processes, several physiological systems are interwoven [2], such as respiratory and cardiovascular systems. In particular, to ensure high efficiency of blood gas exchange under hypoxia, hypoxic ventilation response increases the amount of inhaled oxygen, meanwhile pulmonary vasoconstriction enhances blood flow perfusion [3]. Therefore, HAA is a typical example to explore the concerted responses of multiple physiological systems. In practice, the concerted responses during HAA can be quantified by the correlation analysis of biological phenotypes.
To explore concerted responses to high altitude exposure, we herein applied composite phenotype analysis (CPA) on a longitudinal HAA study ( Supplementary Fig. S1). Application of CPA on four-phase data (plain: Baseline; acute exposure: Acute; chronic exposure: Chronic; back to plain: De-acclimatization) were designed to capture dynamic responses of physiolog-ical systems and their interactions [4,5]. Single phenotype analysis indicated significant fluctuations of multisystem functions during HAA. To integrate the temporal dynamics of phenotypes, we integrated the four-dimensional phenotype across four phases as longitudinal phenotype and dissected their correlation structures. Composite phenotypes were extracted based on clusters of longitudinal phenotypes. Finally, the concerted responses of physiological systems were revealed by correlation network of composite phenotypes.
A total of 33 physiological phenotypes from 822 participants (subjects with cancer, diabetes, and coronary heart disease were not included in this study) were collected at four phases, including Lake Louise Score (LLS) questionnaire, biochemical, hematological and physical aspects ( Supplementary  Fig. S1, Supplementary Methods). The summary statistics showed that most phenotypes changed dramatically across four phases (Supplementary Table S1). For LLS questionnaire aspects, the high prevalence of acute mountain sickness (AMS) has been observed (47.45%) at Acute phase. For biochemical aspects, continuous increased aspartate aminotransferase (AST) was found, suggesting potential liver injury. At Chronic phase, both increased blood urea nitrogen (BUN) and serum creatinine (CREA) suggested the progressive decrease in glomerular filtration rate (GFR) and potential kidney function damage. For hematological aspects, increased red blood cell count (RBC), platelet count (PLT) and white blood cell count (WBC) showed that the hematopoietic system could be activated. For physical aspects, increased heart rate (HR), diastolic blood pressure (DBP) and systolic blood pressure (SBP) reflected that the heart was overloaded with increasing cardiac output. Thus, multiple physiological systems were dynamically changed during the process, and these systems should be taken into account for altitude acclimatization. However, it was unclear how the physiological systems interacted with each other during altitude acclimatization, and it was essential to dissect the relationships among these physiological systems.
Given the plasticity of phenotypes, the correlations of phenotypes were retained across four phases (Supplementary  longitudinal phenotype was performed to construct phenotype correlation network (Fig. 1A). To identify the longitudinal single phenotype structure, we then applied a fast-greedy clustering approach on 33 longitudinal single phenotypes at four phases, and finally obtained nine clusters (Fig. 1B).
The nine composite phenotypes were defined as the set/combination of clustered single phenotypes (Table 1), which were highly consistent with the prior knowledge of the physiological classification. Therefore, these composite phenotypes were accordingly named by physiological knowledge labels (Table 1). Especially, for physical aspects, oxygen saturation (SPO 2 ), heart rate (HR) and blood pressure (SBP, DBP) formed composite phenotype: Circulation. At Baseline phase, there are no strong correlations between SPO 2 , HR and blood pressure (SBP, DBP) under normoxia (Supplementary Fig. S3A and Supplementary Fig. S3B). However, at Acute phase, SPO 2 was correlated with HR, and both HR and SPO 2 were correlated with blood pressure (Fig. 1A and B). Therefore, it is reasonable to take these circulation related phenotypes as one composite phenotype.
To investigate concerted responses among these composite phonotypes, we further adopted CCA to evaluate the correlations between composite phenotypes (Supplementary Table S1 and Fig. 1C and D). Compared with other dimensionality reduction approaches (PCA, etc.), CPA depicted more comprehensive and informative relationships between physiological systems (Supplementary Fig. S4 and Supplementary Methods). Comparing with the network of longitudinal single phenotypes, we found three new associations including the ones between Red Blood Cell and Temperature (rho = 0.365, P-value = 5.26E-09), between White Blood Cell and Liver (rho = 0.287, P-value = 2.43E-05), and between Kidney and Circulation (rho = 0.277, P-value = 3.67E-04) in the network of composite phenotypes. To further explore these three correlations, we constructed composite phenotype networks at each phase separately (Supplementary Fig. S5). The correlation between Kidney and Circulation was replicated at both Baseline and Acute phases, and the correlation between Red Blood Cell and Temperature was also found at Acute phase. However, without CPA, the correlation between White Blood Cell and Liver could not be observed at any phase ( Supplementary  Fig. S5).
Among all human physiological systems, the kidney played a crucial role in regulating body fluids, electrolyte and acid-base homeostasis after acute hypoxia exposure. And Sherpas demonstrated a larger plasma volume than Andeans, resulting in a comparable total blood volume at a lower hemoglobin concentration [6]. Thus, the Kidney and Circulation composite phenotype may be correlated through volume regulation in altitude acclimatization. It was also reported that hemoglobin dynamics [7] in Red Blood Cell were correlated with Temperature. And higher levels of White Blood Cell counts are associated with Liver diseases and enzymes [8]. Based on the above, these three novel associations could give us new insights to understand the potential mechanism of biological processes of altitude acclimatization.
Altitude acclimatization is the physiological process which takes place in the body on exposure to hypoxia at altitude, the most important change is the increase in breathing, and another is the well-known increase in hemoglobin concentration in the blood. As for altitude adaption, Tibetans have higher ventilation, higher oxygen saturation, lower pulmonary artery pressure and lower hemoglobin concentration compared to Han Chinese [9].
Although we collected multiple phenotypes covering the main physiological systems, considering the complexity of specific physiological systems, such as lung, the number of measured phenotypes were still limited. The phenotypes of individuals in this study were collected in the field and the total sample size was not very large, which may limit extrapolation of our results. To achieve more comprehensive analyses of altitude ac-climatization, more frequent data collection would provide more detailed information instead of just four time-points. Molecular phenotypes like epigenome, transcriptome, proteome, metabolome, metagenome and other omics could also provide more valuable information of different physiological systems of altitude acclimatization.
In summary, we applied CPA to reveal potential concerted responses of physiological systems to high altitude exposure. CPA can be considered as a general approach of system biology and phenomics research, especially in large-scale longitudinal cohort studies [10].

SUPPLEMENTARY DATA
Supplementary data are available at NSR online.